ZEO client cache not working properly with beforestorage
If I set my beforestorage to a reasonably current time, e.g. "now", I expect to get reasonably good performance from a ZEO client cache. Instead, I find that beforestorage: 1) asks the cache to ``loadBefore``; 2) but that method only looks for "noncurrent" entries; 3) meanwhile, my cache only has "current" records. I'm not sure if the problem is that the cache shouldn't be setting the loaded oids as "current", or if the ``loadBefore`` method should also accept (some) "current" entries. \malthe
I take issue with "not working properly". Before storage works correctly with ZEO. On Tue, Oct 25, 2011 at 5:43 AM, Malthe Borch <mborch@gmail.com> wrote:
If I set my beforestorage to a reasonably current time, e.g. "now", I expect to get reasonably good performance from a ZEO client cache.
Instead, I find that beforestorage:
1) asks the cache to ``loadBefore``;
This is absolutely the right thing to do.
2) but that method only looks for "noncurrent" entries; 3) meanwhile, my cache only has "current" records.
I'm not sure if the problem is that the cache shouldn't be setting the loaded oids as "current", or if the ``loadBefore`` method should also accept (some) "current" entries.
I agree that the ZEO cache implementation should check for current records with tids < the before time. This has potential implications beyond beforestorage. Would you mind creating an issue: https://bugs.launchpad.net/zodb/+filebug Of course, patches including tests are appreciated. Jim -- Jim Fulton http://www.linkedin.com/in/jimfulton
On 25 October 2011 16:50, Jim Fulton <jim@zope.com> wrote:
Would you mind creating an issue:
Definitely. Filed at: https://bugs.launchpad.net/zodb/+bug/881493
Of course, patches including tests are appreciated.
I'll try and see if I can understand what might be the right thing to expect and write a test case for that. Thanks, \malthe
On 25 October 2011 17:53, Malthe Borch <mborch@gmail.com> wrote:
I'll try and see if I can understand what might be the right thing to expect and write a test case for that.
Following up here on the list to get some feedback on the patch I submitted concerning this issue: https://bugs.launchpad.net/zodb/+bug/881493 The patch includes a test case and necessary code modifications. If someone could review the test case included in the patch, I'd appreciate it. \malthe
On Mon, Nov 7, 2011 at 3:41 AM, Malthe Borch <mborch@gmail.com> wrote:
On 25 October 2011 17:53, Malthe Borch <mborch@gmail.com> wrote:
I'll try and see if I can understand what might be the right thing to expect and write a test case for that.
Following up here on the list to get some feedback on the patch I submitted concerning this issue:
https://bugs.launchpad.net/zodb/+bug/881493
The patch includes a test case and necessary code modifications.
If someone could review the test case included in the patch, I'd appreciate it.
Thanks. I'll review it. Not sure when :), but I'll definitely put this on my to-do list for 3.11. Jim -- Jim Fulton http://www.linkedin.com/in/jimfulton
participants (2)
-
Jim Fulton -
Malthe Borch